Published Paper Details:

: A number of methods for altering faces in films have been effectively created and made publicly accessible in recent years (e.g., Face Swap, deepfake, etc.). Using these technologies, it is possible to facilitate face video modifications with inaccurate results. It is employable in almost all fields. However, the overemphasis of all technologies is fatal prone to have a certain effect in society which may be negative (e.g., fake news, and cyberbullying revenge porn). It is therefore important to be able to tell whether a person's face in a video has been modified, subjectively. To be able to address the problem of deepfake videos, we focus on the problem of face alteration detection in video sequences. In particular, we focus on the ensembles of several Convolutional Neural Network (CNN) models that have been developed.The proposed methodology attains these objectives through the use of attention layers and data training powerful models derived from a base network, EfficientNetB4. By using two publicly available datasets and combining over 119,000 videos, we show how to be able to address these bezier curves, Detecting face alteration is a crucial field of computer vision remains a challenging task in most scenarios, but we demonstrate that in our case, the combined networks approach highly improves the results.